High-speed product appearance inspection is crucial for modern automated industrial production such as bottle caps. The main detection method is to capture the image of bottle caps on the high-speed conveyor belt by industrial cameras and send them to the server through edge devices for various analyses. In order to improve production efficiency, it is necessary to increase the inspection rate of bottle caps. However, the transmission rate and huge data throughput of traditional inspection methods limit the inspection rate. Because of the application requirements for improving bottle cap detection efficiency, this paper proposes a hardware acceleration design based on edge devices to improve the bottle cap detection rate significantly. In this paper, an image processing module based on FPGA is designed as the edge device, improving algorithm execution speed through pipeline processing. It realizes the edge detection of the bottle cap and the fast detection of the front or back states and can send the instructions to the actuator to correct it in time when the back bottle cap is detected. It also realizes the positioning of the bottle cap area and cuts the image. Thereby, the amount of data sent to the server is significantly reduced. We have done both functional simulation and hardware implementation. Comparing with the pure software solution, the proposed design reduces the execution time of the algorithm from 16 ms to 3.07 ms, which achieves a more than four times rate increase. The amount of data that needs to be transmitted to the server per second is reduced from 7200 Kb is reduced to 25 Kb, which reduces the transmission capacity and server cache space by more than 280 times compared to the original image. In this paper, the hardware acceleration design of edge devices and the positioning and cropping of original images can greatly reduce the transmission pressure, data calculation pressure, and buffer space requirements of the central server and improve the detection rate. This design can not only be used for bottle cap inspection but also for various machine vision fields, such as defect detection of other workpieces.